Calibration of the ion microprobe for the quantitative determination of

Rosenstiel School ofMarine and Atmospheric Science, Marine Geology and Geophysics, University ... Although the ion microprobe has been used for geolog...
0 downloads 0 Views 781KB Size
Anal. ch8m. 1990, 62, 722-728

722

Calibration of the Ion Microprobe for the Quantitative Determination of Strontium, Iron, Manganese, and Magnesium in Carbonate Minerals Peter K. Swart

Rosenstiel School of Marine and Atmospheric Science, Marine Geology and Geophysics, University of Miami, Miami, Florida 33149 The neceeoary operam c o "have boen determined for the measurement of trace concentratkns of Mg, Mn, Fe, and SI In calcite and ddomlte uslng a Cameca 3 ion microprobe wlth the energy fllterlng method. With this technlque, the presence of Interferences can be ascertained by examlnlng the devlatlon of the measured lsotoplc ratios of an element from their natural values as a function of the kinetlc energy acquired during the sputterlng process. With the energy 111tering method It is possible, uslng an energy otfset of -50 V, to reduce Interferencesfrom other spedes at masses 24,54, and 88 for Mg, Fe, and Sr. Although Mn has only one stable isotope, It is belleved that a sbnlar voltage offset may be suItable for thls element. By use of these condltlons, callbratlon llnes have been estaMished for Lug, Fe, Mn, and Sr emproylng 11 caldte and ddomite standards. It Is Whaled that for carbonate materlals, concentrations of Mn, Mg, and Sr can be quantitatively measured as low as 1 ppm and Fe as low as 10 ppm. No significant differences In calibration were ofmewed between calcite and dokmlte. The sbpes a d intercepts of the relathsh!ps were found to remaln constant within error over periods of 3 years.

INTRODUCTION Although the ion microprobe has been used for geological and metallurgical problems for a considerable period of time, its application to sedimentary rocks has been limited. In particular the analysis of carbonate minerals has been, until recent studies by Mason (1) and Veizer et al. ( 2 ) ,virtually nonexistent. This investigation reports the analytical conditions necessary to achieve the high resolution and sensitivity necessary for trace element analyses of dolomites and calcites using an energy filtering method previously applied to silicates and sulfides (3, 4). The secondary ion spectrum of any mineral is extremely complex as a result of the presence of a multitude of molecular ion species produced by interaction between the ion beam, the sample, and gases in the background of the mass spectrometer. This process produces dimers, oxides, and hydrides which overlap the various ionic species of interest. For example, the mass spectra shown in Figure 1 are produced through the interaction of an oxygen ion beam with calcite and dolomite samples. Species occurring at major masses have been tentatively identified (Table I). In calcite, the main contribution to mass 40 can be considered to be derived from 40Ca+,but in dolomite as a result of a large concentration of Mg, the species 24MgOt, is also present. In order to correct or remove these interferences and thereby obtain meaningful concentration data, three different approaches, spectral stripping, increased mass resolution, and energy filtering have been used by various workers. The validity of these methods has been discussed by Shimizu and Hart ( 4 ) . Spectral Stripping. This technique, used by Mason ( I ) and Veizer et al. (2), involves the estimation of the contribution of an ionic or molecular species to one particular mass of

Table I. Possible Interferences Occurring at Masses of Interest in This Study

mass

species

%Mg+,%a2+, 23NaH+,12C2t 25Mgt, 24MgHt, "CZH' 26 26Mg+, 24MgHzt, 25MgHt 40 "at NMgl6o+ 41 40CaH+, 24Mg160H+ 26M 16 g o+ WaHH+, =Mgl60Ht 42 42Cat, =Mg O+,2sMg160+, %Sr2+ 43 43Cat, '%aH+, %Mgl80H+,26Mg160H+, 44 %a+ 26Mg180+, %r2+ 46 &Ca+' 54 MFe+, 42Ca12C+ 40ca14N+ 2sMg16012C+ 55 55Mn+,"FeH+,'43Ca12Ct:42Ca12CH+, "Cal4NH+, 26Mg16012CH+,42Ca12CH+ 56 %Fe+,55MnHt, r°Ca160+ UCa12C+,4zCa14NN+ 57 57Fet, 56FeH+,40Ca160H:, UCa12CHt,42Ca14NHt 86 %Sr+,43Ca2t 87 s7Srt, %rH+, &Ca2H+ 88 %Srt,%H+, "Cazt 24 25

interest by measuring related masses. This method was considered to be inadequate (3) because there was no unequivocal method of identifying the ions present or predicting the composition of the sample being analyzed. For example, in the case of @%r+, abundances were corrected for the contribution of @Ca,+ by analyzing mass 80 (1,2). Assuming that mass 80 is composed primarily of 40Ca2+,a contribution for 44Ca2+can be made by subtracting the appropriate isotopic ratio from mass 88. This type of correction will naturally vary according to the chemical composition of the carbonate being analyzed. In particular dolomites will have a large contribution to mass 80 from 40Ca24Mg0+,and therefore separate Calibration curves must be developed for Carbonates with variable Mg concentrations. In addition to taking extra time, counting errors are compounded, as the final ion yield is calculated by subtracting a fraction of the mass 80 yield from that measured a t 88. For other elements, this technique becomes more complicated and subject to more assumptions and uncertainties. Increased Mass Resolution. Increasing the mass resolution of the ion probe allows isotopes to be separated by virtue of their small differences in mass. With this approach %az+, which has an atomic mass of 87.910962, can be separated from %Sr+(mass = 87.905 619). In this particular example, a mass resolution of >16000 would be required. The mass resolution necessary to separate other possible contaminating molecular species that might be encountered in carbonates can be calculated and is tabulated by Mason (1). Necessary resolutions generally exceed 1000. Although this approach has been successfully used with silicates and sulfides (5,6), ion yields from carbonates are significantly lower, so that increasing the resolution of the ion probe decreases the sensitivity to such an extent that in order to obtain a sufficient number of counts, extended counting periods become necessary. This further leads to problems of machine instability over long time periods necessary for counting and, therefore, this approach cannot

0003-2700/90/0362-0722$02.50/00 1990 American

Chemical Society

ANALYTICAL CHEMISTRY, VOL. 62, NO. 7, APRIL 1, 1990

723

CALCITE

lb

IO

0

io

8

DOLOMITE IU

571

10:

O I

'

So

'

SO

70

so

110

170

BO

Flgure 1. Mass spectra produced through the interaction of an oxygen beam with a calcite (A) and dolomite (B) sample. Major molecular and ionic species responsible for the major peaks have been tentattvely identified. For identification of possible species at these masses, see Table 1.

be considered practical for routine elemental analyses in carbonate materials a t this time. Energy Filtering. The final method, which has been utilized in this study, is that of energy filtering as used by Shimizu et al. (31, and first suggested by Herzog et al. (7). The energy-filtering technique can be employed on instruments in which the ion beams form a crossover of rays between the electrostatic and magnetic analyzers. Therefore, the Mattauch-Herzog design as used in the ARL and AE1 probes operated by the University of Chicago and University of Cambridge and reported in the studies of Mason ( I ) and Veizer et al. (2) on carbonates cannot utilize this approach. This design is utilized in the Cameca IMS 3f used in this study. This method has been extensively used for the measurement of trace elements and isotopic ratios in silicates and sulfides (3, 4, 8, 9) but, as yet, not for carbonates. The rationale behind this technique is that different interfering molecular species attain different ranges of kinetic energies during the sputtering process. Oxides (CaO+) and dimers (Caz+)tend to have a much narrower distribution of energies than free ions (Ca+),and therefore with an altered accelerating voltage (using a voltage offset), a segment of the energy spectrum can be examined, which contains proportionally less of these interfering species. The energy spectra of different free, molecular, and complex ions produced in a calcite matrix are shown in Figure 2. Notice in that in Figure 2b the intensity of the ion beam at mass 56 primarily resulting from 40CaO+decreases much more rapidly than if the beam had been composed primarily of single ion such as 56Fe+. Dimers such Ca2+ show similar behavior, while hydrides 40CaH+and doubly charged ions W a 2 + do not exhibit any significantly different behavior compared to metallic ions (Figure 2a). The energy-filtering approach can be tested by examining the isotopic composition of an element as measured with the ion microprobe and comparing it to its correct natural abundance. As interfering species are reduced, the correct natural abundance should be attained. Such measurements have been made on all elements of interest, and the results

are presented in the following section. EXPERIMENTAL SECTION Operating Conditions. The appropriate instrumental conditions were determined by plotting the isotopic ratios of elements against voltage offset; the voltage offset enables ions with a range of kinetic energies to be examined. A voltage offset of -50 V was able to virtually eliminate interferences for all elements with the exception of SBFe(see later discussion). All analyses could therefore be completed at one location on the specimen within a relatively short period of time. For all samples, polished thin sections were prepared and coated with gold. Although only one sample could be mounted at a time, it was possible to prepare standards with several samples on one thin section. Comparisons could then be made between standards analymd with and without letting the specimen chamber up to atmosphere. A primary ion beam of 0-ions was used with a net energy of 13.2 keV (sum of the primary and secondary ion beam potentials). The secondary ion beam was focused before each analysis to give a spot size of between 10 and 15 wm, thereby giving a mass resolution of between 600 and 700 amu. If the beam was improperly focused, then it was possible to produce an ion beam with different energy characteristics. Consequently proper focusing of the beam was an essential prerequisite. After being focused, the beam was left for 15 min on the sample for count rates to stabilize. The stabilization time was determined by measuring intensities for massea 40 and 24 on the sample with respect to time. At the start of a sequence of analyses, each mass of interest was centered by the computer at -50 V offset. Centering the peaks at 0 V frequently led to misidentification of the correct peak position. Counting times for each element varied according to its abundance, but did not exceed 30 s. After background correction, a sequence of five consecutive analyses were averaged and the standard deviations determined. The sample was then moved to a new position and the sequence started again. Although several models have been proposed for obtaining concentration data from intensities of secondary ions (10-12), in this study an empirical approach has been used. In this technique intensities are taken relative to Wa+, as variations in such ratios are relatively insensitive to surface conditions. Materials. The carbonate materials used for standards were obtained from various sources. Three of the calcites (ISC, NCC,

724

ANALYTICAL CHEMISTRY, VOL. 62, NO. 7, APRIL 1, 1990

Table 11. Concentration of Trace Elements of Samples and Relative Intensities Measured in This Investigation

LFC" ISC" NCCn WCb WMb GMC'

TCC' HPD' M85' D100057c C36321'

Mg

Mn

Fe

Sr

723 (128,5) 781 (90,5) 2617 (243,6) 120429 118344 920 746 2059 120104 119480d nm

434 (44,6) 724 (75,6) 1404 (140,7) 432 (44,4) 651 (15,31) 524 516 112451 2029 111 777

5062 (377,6) 1006 (109,6) 14659 (2016,7) 2865 (179,4) 3925 (378,3) 336 2478 90 12476 721 105

480 (59,5) 114 (24,5) 909 (218,6) 69 (25,4) 76 (27,3) 201 897 1983 297 211 188

"Analyses provided by A. E. Dickson (University of Cambridge), AMOCO (ICP, Dr. Fisher), and University of Miami (AA). *Analyses from this paper and AMOCO. 'Analyses from this paper. Figures in parentheses refer to standard deviation and number of analyses. dData for Mg from Jarosewhich and Macintyre (13). Table 111. Relative Intensities of Ion Measured on Samples Listed in Table 11, Measured in 1988

LFC ISC NCC

wc

WM

GMC TCC HPD M85 D 10057 C 36321

24

std dev

55

std dev

54

std dev

88

std dev

8.23-0.4 8.43-04 3.53-03 1.63-01 2.73-01 1.03-03 7.83-04 1.63-03 2.03-01 3.33-01 1.33-04

2.63-04 1.23-04 4.43-04 7.63-02 5.03-03 6.43-04 2.03-04 7.83-04 5.93-02 2.23-02 5.83-05

2.33-04 2.83-04 7.33-04 2.53-04 4.23-04 2.63-04 2.33-04 5.63-03 3.63-03 4.53-05 4.43-04

5.93-05 3.23-05 6.93-05 3.13-05 4.83-05 6.73-05 6.33-05 3.13-03 3.63-04 9.33-05 7.33-05

7.93-05 1.73-05 2.73-04 6.63-05 6.53-05 1.03-05 5.53-05 7.93-06 6.43-04 2.83-05 7.83-06

2.53-05 1.2345 2.93-05 8.93-06 1.43-05 6.53-06 1.3345 2.43-06 5.03-05 3.53-06 2.23-06

4.23-04 1.33-04 8.43-04 1.63-04 1.03-04 2.03-04 8.73-04 2.33-03 3.53-04 4.43-04 2.93-04

5.53-05 2.33-05 1.73-04 3.73-04 4.23-05 1.53-04 1.7344 2.03-03 1.33-04 3.53-05 1.63-06

Table IV

Variation in the Relative Intensity of Seven of the Standards Measured in January 1987

LFC ISC NCC

wc

WM D 10057 C 36321

24

std dev

55

std dev

54

std dev

88

std dev

1.03-03 9.1E-04 3.03-03 3.23-01 3.73-01 3-43-01 2.23-04

1.63-04 1.33-04 6.43-04 2.33-02 5.33-02 1.39-02 5.73-05

1.7344 2.73-04 6.23-04 2.1344 3.33-04 1.13-04 4.93-04

3.23-05 8.0345 1.23-04 7.63-04 8.33-05 3.1346 1.7345

8.83-05 2.23-05 2.13-04 8.63-05 7.63-05 2.33-05 2.43-05

1.5345 3.13-06 4.03-04 4.33-05 2.63-05 9.03-06 1.13-05

3.83-04 1.33-04 8.23-04 1.3344 9.13-05 3.93-04 2.43-04

1.5344 7.0345 3.2344 1.5345 3.13-05 1.23-04 1.53-05

Variation in the Relative Intensity of Three Standards Measured in January 1986

LFC ISC NCC

24

std dev

55

std dev

54

std dev

88

std dev

1.13-03 1.39-03 3.53-03

3.33-04 5.43-04 3.83-04

1.53-04 3.83-04 5.03-04

2.53-05 8.03-05 1.03-04

5.13-05 2.93-05 4.63-04

1.43-05 98.03-07 6.43-04

4.53-04 2.53-04 9.93-04

2-13-04 1.23-05 1.83-04

and LFC), provided by A. E. Dickson (University of Cambridge, U.K.), are identical with those reported by Mason (1). We have also included two other analyses on these samples, one measured by using atomic absorption (AA) techniques a t the Rosenstiel School of Marine and Atmospheric Sciences, University of Miami, and the other by inductively coupled plasma (ICP) a t AMOCO Research (Table 11). In addition, six new samples, three calcites, HPD, LCC, and FCC, and three dolomites, WC, WM, and M85, as well as the Smithsonian calcite and dolomite standards, C136321 and D 10087,were analyzed (13). Wet chemical analyses on these eight samples were obtained from AMOCO and from our own analyses presented here (Table 11). Data presented in the following section consist of an average of 20 spots on each standard. Error bars represent i l standard deviation. During the collection of data the standard deviation of an individual analysis can be estimated by the approximation l/drz*mean (n = number of counts) (Table 111). A best fit line was then calculated for each element and the standard error of the slope and constant determined. The equations are shown in and 11. In addition, data for five of the standards, Figures 5,6,9, NCC, ISC, LFC, WC, WM, C36321,and D10057,were obtained

over a period of two years and data for three standards, NCC, ISC, and LFC, were obtained over three years. These values are included in order to ascertain the reproducibility of the standards (Table IV). In order to utilize our measurements to calculate concentrations of trace elements, an empirical approach was used in which the ratio of the mass of interest relative to calcium was correlated with the concentration of the trace element in the carbonate. Mass 40 was used as a reference mass, rather than mass 42, as in previous studies of carbonates (1, 2). These workers used mass 42 because the intensity of mass 40 under the instrumental conditions they used was too great and could possibly damage the collector. Although similar problems were found in this study a t 0 voltage offset, intensities were sufficiently reduced at the offsets eventually used that no problems were encountered. Veizer et al. (2) also reported significant problems from interferences of 28SiO+a t mass 42 as result of silicate inclusions in their carbonates. No such problems were encountered in this study, both as a result of the apparent absence of such inclusions in our samples and also because the energy filtering method was able to reduce such interferences to an acceptable level (Figure 3).

ANALYTICAL CHEMISTRY, VOL. 62, NO. 7, APRIL 1, 1990

A

C

O

U

M85\A

~

-

-10'

0.022 0 022

1

0

p.

0.02

725

1

-

NCC

3

0.009

/

-

0.006-

' + +

+ +

Figure 3. Change in mass 44/(40 42 44) and 42/(40 42 44) ratio with respect to increasing energy offset for a dolomite (M85) and a calclte (NCC). The horizontal line represents the natural abundance. Vertical bars are equal to 1 standard deviation.

a 2 %+

g

0.13

1

0.12

1

I

0.11 0.1

0.08

4

VOLTAQE OFFSET

- 125

-75

-25

0

1

Figure 2. Energy spectra for various ionic and molecular species produced by the reaction of an 0-beam and a calcite sample. I n part A masses 40 (%a2+) and 41 ('OCaH') are shown relative to mass 'Oca+. These species show no differences In their energy spectra compared to %a+. In part B "%a2+and %a''O+ are shown relathre to %a+. Note that at Increased offset voltages the ratio of simple ionic specles to more complex molecular species increases.

0.07 0'14

~

I

RESULTS Calcium. Although calcium is the most abundant element in all the samples analyzed, significant contributions from 2eMg160+and urMg180+to mass 42 are possible in Mg-rich samples such as dolomite. Additional interferences may result from 12c1602+, @CaHH+,and "CaHH+. The presence of these interferences can be clearly seen in Figure 3. With an offset of -20 V,the 44/(40 42 44) and 42/(40 + 42 + 44) ratios can be reduced to within one standard deviation of the natural abundance of these isotopes. Magnesium. Magnesium has three stable isotopes, 24,25, and 26, which have natural abundances of 78.99,10.00, and 11.01%. The percentages of mass 25 and 26 are plotted in Figure 4 for a dolomite standard (M85) and a calcite (NCC)

+ +

1

0.07 0.08-100

-80

-80

-40

-20

0

Offset (v)

+ +

Figure 4. Change in the mass 25/(24 25 26) and mass 26/(24 25 26) ratio with respect to increasingenergy offset for a dolomne and calcite sample as in Figwe 4. The horizontal b e represents natural abundance. Vertical bars are equal to 1 standard deviation.

+ +

as a function of voltage offset of the accelerating voltage. From these data it can be seen that at 0 voltage offset there is

1

t

1

i

I TOE-3

i 10E-4

~

10

I

,,,' '

,

, , , , , ,,, 100

, , , , , ,,,, 1wO

, , , , ,,,,,

,

, I

1

,

,,,,,,

,

, ,

Figure 5. Calibration line between mass 24/40ratio and concentration of magnesium in the bulk sample. The error bars represent fl

standard deviation. The values shown in brackets are one standard error. The sdkl circle markers represent analyses of standards during 1987. Data for these are shown in Table IIIa. significant deviation from the correct isotope ratio. Such deviation takes the form of an excess of mass 25 and disappears with an offset of -50 V. Although energy filtering is generally not effective a t removing hydride interferences, in this case the most probable interference at mass 25 is 24MgH+. Other possible interferences at these masses are listed in Table

I. The concentration of Mg ranged from 12.04% in the dolomites to 723 ppm in LFC. As a result of the large differences in Mg concentration between dolomite and calcite, the correlation line in the calcite samples has been extrapolated to the dolomites. Although these dolomites lie within error on this line (Figure 5), the dolomites generally gave inconsistent mass 24/40 ratios. At the present time results on Mg-rich samples such as dolomites do not appear to be consistent enough to allow small differences in Mg content to be determined. Manganese. Manganese is monoisotopic and therefore it is impossible using this approach to determine the presence of interferences. Possible interferences are shown in Table I. In the study of Veizer et al. (2),43Ca12C+was considered to the most probable interfering ion, while Mason (1) suggested, 25Mg18012C+ and 26Mg1s013C+.Veizer et al. (2) suggested that 43Ca12C+would contribute approximately 1 ppm Mn. Mason (1) concluded that the effect for MgOC+ species was approximately 2 ppm per loo0 ppm Mg and therefore only important in dolomites. Considering that both of these studies ( I , 2) were conducted a t lower mass resolutions than this investigation and the finding in this study that significant interferences can be present even for the major elements, it is unlikely that quantitative concentration data can be obtained by using the approach of Veizer et al. (2) and Mason (1). Furthermore it is probable that major interferences not identified by these workers are also present. These include hydrides such as 26Mg16012CH+ and 42Ca12CH+. The concentrations of the standards used in this study range from 111 to 12476 ppm (Table 11). With the exception of two samples of dolomite (M85 and HPD), all fall within one standard deviation of the best fit line (Figure 6). The line is approximately linear over 2 orders of magnitude and therefore can be extrapolated to concentrations of 10 ppm Mn, corresponding to a 55/40 ratio of 1 X The percent error estimated from counting statistics for the standard with the lowest concentration of Mn (111 ppm in D10057) is approximately 4%. Iron. The spectrum of iron is dominated by 40CaO+and 40CaOH+. Even with a voltage offset of -90 V, the isotopic

,1,,,)

, , , , ,,,,,

1wO Mn (ppm)

100

10

loo00

Mg ( w m )

I

HPD

I

'e

,

10E-5

, , , ,,

,

,

,

,

,,,,,I

loo00

Figwe 6. Calibration line for mass 55/40and the concentration of Mn. The error bars represent f l standard deviation. The values shown

in parentheses are one standard error. The solid circular markers represent analyses of standards during 1987. Data for these are shown in Table IIIa. 1

/-

h

r;;

$ .01 In

+

. d

ln

v

b In

I

,001

i ,001

4

u I

I

-100

-80

-60

-40

-20

0

Offset (v)

Figure 7. Mass 54454

+ 56 + 57)and 57454 + 56 + 57)ratio with

respect to energy offset; horizontal line represents the natural abundance. Vertical bars are equal to 1 standard deviation. Note that at -50 V offset there is still a large difference between measured values and natural abundances. ratios of Fe are outside one standard deviation of their natural value (Figure 7). However, even at these energy offsets, the intensity of the mass 56 peak exhibits a strong correlation with that of the mass 40 peak suggesting that unresolvable differences were still present. As a result of problems with these interferences in previous studies (1,2), it was necessary to measure intensities at mass 54. This places a severe limitation of the detection of iron compared to other elements as 54Fe represents only 5.8% of the total abundance of Fe. Sensitivity is therefore reduced by between 10 and 20 times compared to Mn. In addition, the presence of high carbon backgrounds in some instruments may result in the production of species such as 40Ca12CH2+ interfering at mass 54, thereby further

ANALYTICAL CHEMISTRY, VOL. 62,NO. 7,APRIL 1, 1990

727

0.2

0.16

10E-4

1

4

04 0.9,

10E-6

.:

, 100

,,,,,,,, 1000

, ,,,,,,,, 10000

, ,,,,,,,,

, ,,,,,

07

100000

Fe (ppm)

I

I

I

I

I

,

I

I

I

I I

i

Figure 8. Calibration line between mass 54/40 ratio and the concentration of Fe in ppm measured in bulk samples. The error bars represent fl standard deviation. The values shown in brackets are one standard error. The solM circular markers represent analyses of standards during 1987. Data for these are shown in Table IIIa. + .loo

,

, -80

,

, .BO

,

, -40

,

, -20

,

I 0

Offset (v)

+ +

+ +

Figure 10. Change in mass 86/(86 87 88)and 87/(86 87 88) ratio with respect to increasing energy offset; horizontal line represents the natural abundance. Vertical bars are equal to 1 standard deviation. 1OE-3

I 1000 10000

10

10E4

100

Fe ( P P ~ )

F@re D. Relationship between mass 56/40ratio and Fe concentration for six samples at -90 V offset. The relationship between 54/40and Fe concentration (Figure 9)is shown for comparison. Note the absence of significant Correlation between the 56/40ratio and Fe content.

reducing sensitivity for this element. The concentrations of iron in the standards used in this study range from 90 to 14 500 ppm. The log of the 54/40 ratio plotted against the log of the iron concentration for all samples, with the exception of M85, falls close to one standard deviation from the best-fit line (Figure 8). However, the 56/40 ratio shows little relationship to the concentration of iron and is directly correlatable to the intensity of mass 40 (Figure 9). It is possible that with the use of greater voltage offsets correct 56/(54 + 56 + 57) isotopic ratio could be obtained; however, the use of large offsets reduces the intensity of the ion beam appreciably and therefore reduces the sensitivity and detection limits of other trace elements. The results therefore support the conclusions of Mason ( I ) and Veizer et al. (2) that since interferences a t mass 56 cannot be easily resolved, mass 54 is the most suitable isotope for measuring concentrations of Fe. However, the standard deviation estimated from total counts is high in samples with low Fe concentrations. For example, in HPD the total number of counts used to calculate the mass 54/40 ratio is approximately 100, corresponding to an error of 10%. Strontium. The principal interference at mass 88, %az+, can be removed with an energy offset of -20 V (Figure 10). At offsets lower than this value, the 87/(86 + 87 + 88) and 88/(86 + 87 + 88) ratios of the two standards investigated remained constant within analytical error. The strontium concentration of the samples investigated in this study ranged from 10 000 to approximately 100 ppm. The log of the mass 88/40 ratio plotted against the log of the

9 . m

10E-5

I

2

Figure 11. Calibration line between mass 88/40ratio and the concentration of Sr in ppm measured in bulk samples. The error bars represent f 1 standard deviation. The values shown in parentheses are one standard error. The soli circular markers represent analyses of standards during 1987. Data for these are shown in Table IV.

strontium concentration for all samples, including dolomites, plotted on a line with a slope of 1.05 (f0.09) and a correlation coefficient of +0.98 (Figure 11). This is in contrast to the results of Veizer et al. (2) who needed to establish separate calibration lines for Sr in calcite and dolomite. Although the lowest Sr standard used was 95 ppm, the linearity of the relationship from loo00 to 100 ppm, enables the extrapolation to 88/40 ratios as low as lo4, equivalent to a concentration of 1 ppm. The percent counting error on the sample with lowest Sr concentration (90 ppm, WC) is approximately 2%. Long-Term Reproducibility. The majority of the data presented in this paper was collected during one session in January 1988. However, over the previous two years, data were collected under similar conditions for five of the standards finally analyzed. These data, shown in Table IV and in Figures 5,6,8, and 11,indicate that with the exception of Mg, similar analyses fall within one standard deviation of each other, suggesting that the slope of the relationships established can be utilized even if all standards were not

ANALYTICAL CHEMISTRY, VOL. 62, NO. 7, APRIL 1, 1990

728

10E-2

-

10E-3

.-0 ?! 0

2 lOE-4 m In u?

I

8

I

54/40(+ lsigma)

10E-6

j 1

,

. 3

,

, , , 5

7

-I 9

11

13

15

17

Transect Posrtion

Figure 12. Results of 18 consecutive spot analyses across sample HPD. Each spot is approximately 20 pm apart. The lines for each

mass indicate the mean ratio plus and minus one standard deviation. The standard deviation has been estimated by using the expression lldn'mean. Hence for a total number of counts of 100, the standard deviation (CT) is 10% of the mean. Note that for both masses 88 and 55, variations are significantly greater than c a n be explained by counting statistics. In the case of mass 54, a 10% variation is able to account for most of the observed variability. analyzed at the same time as the samples. DISCUSSION The data reported here represent the first attempt to apply the energy filtering approach to the quantification of the measurement of trace elements in carbonates, although there are many reports of using this technique in silicates ( 3 , 4 , 8 , 9). Previous studies, which have attempted to use the ion microprobe for the analysis of trace elements ( I , 2) in carbonates, used a spectral stripping method. Although these authors reported the analysis of trace elements at the parts-per-million level, the spectral stripping approach is still subject to a number of uncertainties. In particular at the mass resolutions of 500-600 amu (this study used higher mass resolutions) used previously (I, 2), this investigation has identified the existence of substantial interferences at all masses. Our approach has shown that using the energy filtering method, isotopic abundances can be measured that are statistically the same as natural occurrences. Important interferences can therefore be considered to have been eliminated and relative intensity ratios accurately measured. In addition, our studies on some of the samples indicate that even over periods as long as three years, similar ratios can be obtained on standards. Therefore, if instrumental conditions are similar, the relative ion beam intensities can be related directly to the relative intensities obtained in this investigation. The long-term reproducibility studies are in contrast to those reported by Veizer et al. (2) who showed significant differences between 88/42 ratios measured on separate occasions, differences that were greater than the standard deviation of the measurement. The error involved in the calculation of any calibration between intensity ratios and concentration is dependent upon the compositional homogeneity of the standards analyzed. While there is no definitive way of ascertaining homogeneity before analysis, if sufficient analyses are made on a sample, a true mean intensity for the standard can be measured. Some of the standards analyzed in this investigation proved to be

heterogeneous for certain elements, ISC and GMC for Fe, HPD for Mn, and GMC and HPD for Sr. That the variation in trace element content is in fact true heterogeneity, and not a result of analytical uncertainty, can be verified by comparing the range of ratios with the standard deviation for each analysis estimated from counting statistics. For one of the standards that proved to be particularly heterogeneow (HPD), the ratios of 54/40,55/40, and 88/40 have been plotted plus or minus one standard deviation. These data indicate that the change in ratios of 55/40 and 88/40 are all in excess of what might be expected from counting statistics (Figure 12). In contrast the 54/40 ratio exhibits a greater range of uncertainty as a result of the low count rate and its high standard deviation. Certain of the other standards such as M85 deviated significantly from the best fit line for several elements. Such deviation is best explained in terms of analytical inaccuracy of the wet chemical datta, even though repeated analyses produced similar results. An alternative explanation is that this sample was extremely heterogeneous for these elements. Contamination appears unlikely as according to the best fit line, the concentration of iron and manganese measured by wet chemical means was lower than that calculated by using the ion microprobe data. Typically bulk analyses produce higher concentrations as a consequence of the inability to remove small amounts oxides, clays, and other contaminants trapped between grains. ACKNOWLEDGMENT Thanks are extended to S. Hart and N. Shimizu for use of the ion microprobe at MIT. Technical assistance was provided by V. Vahrenkamp and M. Guzikowski. Calcite standards (ISC, NCC, and LFC) were kindly provided by A. Dickson. J. Jarosewich of the Smithsonian Institute provided sufficient quantities of C36321 and D10057 to allow wet chemical determination of their Sr, Mn, and Fe concentrations. Additional analyses were provided by B. Fisher of AMOCO. S. Burns is thanked for making wet chemical analyses at RSMAS. LITERATURE CITED Mason, R. A. Chem. -1. 1987, 64, 209-224. Veizer, J.; Hinton, R. W.; Clayton, R. N.; Lerman, A.

Chem. Gee/. 1987, 64, 225-237. Shimizu. N.; Semet, M. P.; Aliegre. C. J. Geochim. Cosmochim. Acta 1978. 42, 1312-1334. Shimizu, N.; Hart, S. R. Annu. Rev. Earth Planet. Sci. 1982, 10, 483-526. Steele, I. M.; Hutcheon, I . D.; Smith, S. V. Geol. SOC.Am. Abstr. Programs 1978, 8 , 1119. Hinton, R. W.; Long, J. V. P. Earth Planet. Sci. Left. 1979, 45, 309-325. Herzog, R. F. K.; Poschenrieder, W. P.; Satiewicz. F. 0.Radiat. Eff. 1973. 18, 199-205. Shimizu, N. Phys. Chem Earth 1975, 9 . 655-669. Shimizu, N. Earth Pianet. Sci. Left. 1978, 39, 398-406. Anderson, C. A.; Hhthorne, J. R. Anal. Chem. 1973, 45, 1421-1438. Anderson. C. A. NBS Spec. Pub/. 1975, No. 427, 79-119. Simonds, 0.S.;Baker, J. E.; Evans, C. A. Anal. Chem. 1976, 48, 1341- 1348. Jarosewich, E.; Macintyre, I. G. J . Sediment. Petrol. 1983, 53, 677-678.

RECEIVED for review August 18, 1989. Revised manuscript received January 3, 1990. Accepted January 3, 1990. This research was supported by NSF Grants EAR 84-19359 and 86-07688to P.K.S. Additional funding was supplied by Shell Research B.V.